10 for analyzing a biochip image may detect the camera intrinsic parameter matrix and the distortion coefficient corresponding to the camera, so that the validity of the camera intrinsic parameter matrix and the distortion coefficient can be ensured.

In some embodiments, the original image is a fluorescent image of the biochip in which a biochemical reaction has occurred.

When a biological sample to be detected is loaded on the biochip and a biochemical reaction occurs, a corresponding fluorescent image of the biochip may be acquired by using a specific device. It is understood that different reaction chambers may show the same color and intensity or different colors and intensities in the fluorescent image.

Referring to FIG. 4, in some embodiments, the preprocessed image includes an image having a high-frequency component, and step S1 includes:

- Step S13, performing a Gaussian filtering processing on the biochip image to obtain an image having a low-frequency component; and
- Step S14, subtracting the image having the low-frequency component from the biochip image to obtain the image having the high-frequency component.

Specifically, the step S13 and the step S14 may be implemented by the acquisition module 11. That is, the acquisition module 11 may be configured to perform the Gaussian filtering processing on the biochip image to obtain the image having the low-frequency component, and to subtract the image having the low-frequency component from the biochip image to obtain the image having the high-frequency component.

Therefore, the image having the low-frequency component is obtained by the Gaussian filtering, and then the image having the high-frequency component is obtained by subtracting the low-frequency component in the biochip image, so that the high-frequency filtering is realized, and the problem of uneven fluorescent illumination of the microscopic chip is solved.

Alternatively, in other embodiments, the preprocessed image may not be limited to the above image having the high-frequency component, but may be obtained as a grayscale image, the image having the low-frequency component, an edge detection image, or the like as needed. The grayscale image may be obtained through image graying processing, the image having the low-frequency component may be obtained through low-frequency component extraction processing, and the edge detection image may be obtained through image edge extraction processing. In addition, the preprocessed image may be obtained by using one or more of the above processing methods in a preset order, which is not limited herein.

Referring to FIG. 5, in some embodiments, step S2 includes:

- Step S21, selecting a preset number of detection areas in the preprocessed image;
- Step S22, detecting a center and a radius of each area of interest in each detection area by using a Hough circle transformation (cvHoughCircles); and
- Step S23, forming a circle according to the center and the radius of each area of interest, to determine the area of interest and segment the area of interest.

Specifically, step S21, step S22, and step S23 may be implemented by the correction module 12. That is, the correction module 12 may be configured to select the preset number of detection areas in the preprocessed image; detect the center and the radius of each area of interest in each detection area by using the Hough circle transformation (cvHoughCircles); and form the circle according to the center and the radius of each area of interest, to determine the area of interest and segment the areas of interest.

By taking the area of interest as an area where a reaction chamber is located in the image as an example, when an arrangement of the reaction chambers is detected, it needs to determine a position of each reaction chamber in each detection area. Each reaction chamber is generally circular, so that a center and a radius of each chamber may be detected through the Hough transformation. Further, the position of each reaction chamber is determined according to the center and the radius of each chamber, and then, the segmentation of the reaction chambers may be realized.

In some embodiments, step S21 includes: selecting a corresponding detection area within a predetermined area of the preprocessed image.

The correction module 12 may be configured to select the corresponding detection area within the predetermined area of the preprocessed image.

The predetermined area may be empirically set by a user, or automatically selected through an algorithm. Alternatively, the detection area may also be a randomly selected area in the preprocessed image, and is not limited here.

In some embodiments, each detection area is a rectangular area, and includes at least two rows or at least two columns of areas of interest.

It is understood that when the deflection correction is performed on the image having the high-frequency component to obtain a deflection-corrected image, it is necessary to determine a deflection angle of the image having the high-frequency component. The reaction chambers on the biochip are generally arranged in an array, i.e., the areas of interest are generally arranged in an array, so that the deflection angle of the image (an image deflection angle) may be detected according to an arrangement direction of the chambers. Advantageously, the rectangular area is used for determining a relative deflection angle between a long side direction of the selected detection area and the arrangement direction of the reaction chambers.

Each detection area includes the at least two rows or at least two columns of areas of interest, which may ensure the detection of the arrangement direction of the reaction chambers.

It should be noted that a size of each detection area may be flexibly set according to an area interval of the areas of interest, a radius of each area of interest, or the like, and is not limited herein.

Alternatively, in other embodiments, a shape of each detection area may not be limited to the rectangular shape. Other suitable shapes such as a square, a triangle, a circle, a parallelogram, etc. may be selected as needed, which is not limited herein.

In some embodiments, the preset number of the detection areas selected each time may include a plurality of detection areas. The plurality of detection areas may have different directions, so as to improve the efficiency and accuracy of detecting the deflection angle of the image. For example, the preset number of detection areas selected each time may include nine detection areas.

Referring to FIGS. 6 and 7, in some embodiments, step S2 includes:

- Step S24, performing an expansion processing on the segmented image to connect adjacent areas of interest in a preset direction;
- Step S25, performing a principal component analysis on a contour of the detection area having a maximum contour in the image on which the expansion processing is performed, to obtain a contour direction; and
- Step S26, determining an image deflection angle according to the contour direction, to correct the preprocessed image and to obtain the deflection-corrected image.

Specifically, steps S24, S25 and S26 may be implemented by the correction module 12. That is, the correction module 12 may be configured to perform an expansion processing on the segmented image to connect adjacent areas of interest in a preset direction; perform a principal component analysis on a contour of the detection area having a maximum contour in the image on which the expansion processing is performed, to obtain a contour direction; and determine a deflection angle of the image according to the contour direction, to correct the preprocessed image and to obtain the deflection-corrected image.

In step S24, each segmented area of interest may be expanded in a preset direction, so that a contour of each area of interest extends in the preset direction, and contours of adjacent areas of interest are connected to each other. In one example, the preset direction may be the long side direction of the rectangular detection area. FIG. 7 is a schematic diagram showing contours of all nine detection areas obtained after the expansion process is performed on the areas of interest in the nine detection areas according to the preset direction.

After the contours of the areas of interest are connected together, in step S25, the principal component analysis (PCA) is performed on the contour of the detection area having the maximum contour, to obtain the contour direction. It is understood that the maximum contour is generally formed by connecting the plurality of areas of interest adjacent to each other, so that the obtained contour direction may be used as the arrangement direction of the reaction chambers in one example. In particular, when the plurality of detection areas are included, the PCA principal component analysis may be performed on the contour of one of the plurality of detection areas having the maximum contour, to obtain the contour direction.

Thus, in step S25, the image deflection angle of the biochip image and the preprocessed image may be determined according to the contour direction, and the correction for the deflection angle may be performed on the biochip image and/or the preprocessed image, to obtain the deflection-corrected image.

Thus, in the present disclosure, the detection area having the maximum contour is formed by connecting areas of interest adjacent to each other in a same direction, and then, the principal component analysis (PCA) is used, so that the problem of the detection for the deflection angle can be solved.

Referring to FIG. 8, in some embodiments, step S2 includes:

- Step S27, enlarging the selected area by a preset proportion, to randomly select the preset number of detection areas in the preprocessed image again; and
- Step S28, repeatedly and iteratively detecting the image deflection angle until the image deflection angle is smaller than a preset angle threshold, to obtain the deflection-corrected image.

Specifically, steps S27 and S28 may be implemented by the correction module 12. That is, the correction module 12 may be configured to enlarge the selected area by a preset proportion, to randomly select the detection areas again; and repeatedly and iteratively detect the image deflection angle until the image deflection angle is smaller than a preset angle threshold, to obtain the deflection-corrected image.

Therefore, the image deflection angle is repeatedly and iteratively detected with the detection areas with different sizes, and therefore the accuracy of the image deflection angle can be ensured.

In some embodiments, a range of the preset angle threshold may be determined by the following conditional expression:

cos
⁢

θ

-

(

max
⁢

{

m
,
n

}

-
1

)

·
sin

⁢

θ

>

2
⁢

rad

d
⁢
i
⁢
s
⁢
t

- where θ is the preset angle threshold, dist is the area interval of the areas of interest, rad is an area radius of the area of interest, m is the number of rows of the areas of interest in each detection area, and n is the number of columns of the areas of interest in each detection area.

It should be noted that in some embodiments, when shooting the biochip image, the camera may be precisely aligned by a hardware instrument, so that the deflection-corrected image may be directly determined from the biochip image shot after the precise alignment. In this case, the process of detecting the image deflection angle may be omitted. In other embodiments, a sign may be provided on the biochip itself. After the biochip image is obtained, a relative coordinate system may be constructed by identifying the sign on the biochip, to obtain a deflection angle of the biochip with respect to the camera, so as to obtain the deflection-corrected image by correction.

Alternatively, the correction for angle deflection may not be limited to the above embodiments, and a suitable correction method may be selected as needed, so that the apparatus 10 for analyzing a biochip image may determine that the areas of interest on the biochip satisfy relative positions arranged laterally or longitudinally according to the deflection-corrected image, which is not limited herein.

Referring to FIG. 9, in some embodiments, step S2 includes:

- Step S27′, enlarging the selected area by a preset proportion, to randomly select the preset number of detection areas in the preprocessed image again; and
- Step S28′, repeatedly and iteratively detecting the image deflection angle for a preset number of times, to obtain the deflection-corrected image.

Specifically, steps S27′ and S28′ may be implemented by the correction module 12. That is, the correction module 12 may be configured to enlarge the selected area by the preset proportion, to randomly select the detection areas again; and repeatedly and iteratively detect the image deflection angle for the preset number of times, to obtain the deflection-corrected image.

Therefore, the image deflection angle is repeatedly and iteratively detected for the preset number of times with the detection areas with different sizes, and therefore the accuracy of the image deflection angle can be ensured. In one example, the preset number of times may be preset by a system or set by a user as needed. For example, the preset number of times may be six times.

Referring to FIGS. 10, 11 and 12, in some embodiments, step S3 includes: Step S31, constructing a notch filter; and

- Step S32, filtering the deflection-corrected image by using the notch filter, to obtain an image with enhanced periodic patterns.

Specifically, step S31 and step S32 may be implemented by the processing module 13. That is, the processing module 13 may be configured to construct the notch filter; and filter the deflection-corrected image by using the notch filter, to obtain the image with enhanced periodic patterns.

Therefore, with the periodic pattern prior of the array type biochip, the notch filter is constructed, so that a noise caused by a surface dirt, a sample injection process and a reaction process is weakened to the maximum extent.

In one example, in an amplitude map obtained by transforming the biochip image through the Fourier transform into a frequency domain, as shown in FIG. 11, most of the image information is concentrated in a low-frequency part, so that most of the image noise can be removed by filtering the image information in a central part. The information in a central vertical direction and a central horizontal direction is the periodic pattern information that is most easily filtered out, so that the notch filter constructed in step S31 may be shown in FIG. 11, for removing the image noise in a non-periodic pattern.

The notch filter is used, which can reduce the time and space complexity of an algorithm of the method for analyzing a biochip image of the embodiment of the present disclosure, and have a looser requirement on a performance of a hardware device, thereby reducing the cost and improving the operation efficiency while ensuring the effect.

Referring to FIG. 13, in some embodiments, step S33 includes:

- Step S331, performing a smooth filtering processing on the image with enhanced periodic patterns by using a box filter;
- Step S332, integrating pixel values of the image on which the smooth filtering processing is performed in a horizontal direction and a vertical direction, to obtain a first integral curve in the horizontal direction and a second integral curve in the vertical direction, respectively, and taking a set of minimum points of the first integral curve and the second integral curve to determine a grid spacing line; and
- Step S333, dividing grid areas according to the grid spacing line.

Specifically, steps S331, S332, and S333 may be implemented by the processing module 13. That is, the processing module 13 may be configured to perform a smooth filtering processing on the image with enhanced periodic patterns by using a box filter; integrate pixel values of the image on which the smooth filtering processing is performed in a horizontal direction and a vertical direction, to obtain a first integral curve in the horizontal direction and a second integral curve in the vertical direction, respectively, and take a set of minimum points of the first integral curve and the second integral curve to determine a grid spacing line; and divide grid areas according to the grid spacing line.

Therefore, the smooth filtering processing is performed by using the box filter, so that other noise interference in the image can be weakened, so as to divide grid areas.

In some embodiments, a length or width of an operator of the box filter satisfies the following conditional expression:

⌈

d
⁢
i
⁢
s
⁢
t

2

-
rad

⌉

<
b
<

⌈

d
⁢
i
⁢
s
⁢
t

2

⌉

- where b is the length or width of the operator of the box filter; dist is the area interval of the areas of interest, rad is an area radius of the area of interest.

Referring to FIG. 14, in other embodiments, step S3 includes:

- Step S34, detecting areas of interest in the deflection-corrected image by using a Hough circle transformation; and
- Step S35, forming circles according to the detected areas of interest, and performing an approximation to obtain the grid areas through division.

Specifically, steps S34 and S35 may be implemented by the processing module 13, that is, the processing module 13 may be configured to detect areas of interest in the deflection-corrected image by using a Hough circle transformation; and form circles according to the detected areas of interest, and performing an approximation to obtain the grid areas through division.

That is, for the division of the grid areas, the areas of interest in the deflection-corrected image may be detected by using the Hough circle transformation, and then circles are formed according to the detected areas of interest, and an approximation is performing to obtain the grid areas, thereby achieving the division of the grid areas.

Alternatively, the division of the grid areas may not be limited to the above embodiments, and other division methods may be adopted as needed to divide the grid areas, which are not specifically limited herein.

Referring to FIG. 15, in some embodiments, step S3 includes: